net: stmmac: fix DMA error when resume with specific PHY

When suspend is performed, "Power down" cmd (PWD) is send to PHY BMCR register.
Some phy (RTL8211F) cut RXC clock. And when resume is called we perform
firstly SWR reset (in stmmac_hw_setup) and after with resume PHY (phy_start),
this sequence is wrong because SWR is failed due to RXC off.
This patch consist to resume firstly PHY and after we perform SWR reset

Signed-off-by: Christophe Roullier <christophe.roullier@st.com>

Change-Id: Id7b0c552c34b4e4a407c016fbcbd6e3f7080634d
Reviewed-on: https://gerrit.st.com/c/mpu/oe/st/linux-stm32/+/141914
Reviewed-by: CITOOLS <smet-aci-reviews@lists.codex.cro.st.com>
Reviewed-by: Christophe ROULLIER <christophe.roullier@st.com>
Reviewed-by: CIBUILD <smet-aci-builds@lists.codex.cro.st.com>
Reviewed-by: Alexandre TORGUE <alexandre.torgue@st.com>
Tested-by: Christophe ROULLIER <christophe.roullier@st.com>

Author: Christophe Roullier

drivers/net/ethernet/stmicro/stmmac/stmmac_main.c

@@ -4488,9 +4488,6 @@ int stmmac_suspend(struct device *dev)
 	/* call carrier off first to avoid false dev_watchdog timeouts */
 	netif_carrier_off(ndev);
 
-	if (ndev->phydev)
-		phy_stop(ndev->phydev);
-
 	mutex_lock(&priv->lock);
 
 	netif_device_detach(ndev);
@@ -4516,6 +4513,10 @@ int stmmac_suspend(struct device *dev)
 	priv->oldlink = false;
 	priv->speed = SPEED_UNKNOWN;
 	priv->oldduplex = DUPLEX_UNKNOWN;
+
+	if (ndev->phydev)
+		phy_stop(ndev->phydev);
+
 	return 0;
 }
 EXPORT_SYMBOL_GPL(stmmac_suspend);
@@ -4608,6 +4609,9 @@ int stmmac_resume(struct device *dev)
 		goto init_error;
 	}
 
+	if (ndev->phydev)
+		phy_start(ndev->phydev);
+
 	stmmac_hw_setup(ndev, false);
 	stmmac_init_tx_coalesce(priv);
 	stmmac_set_rx_mode(ndev);
@@ -4618,9 +4622,6 @@ int stmmac_resume(struct device *dev)
 
 	mutex_unlock(&priv->lock);
 
-	if (ndev->phydev)
-		phy_start(ndev->phydev);
-
 	return 0;
 init_error:
 	free_dma_desc_resources(priv);